ut3_save
========

.. py:function:: t3toolbox.backend.ut3_constructors.ut3_save(file, data)

   .. code-block:: python

      def ut3_save(
              file,         # path or open file object to write the .npz to
              data: UT3Data,
      ) -> None:


   Save a uniform Tucker tensor train (2 supercores + 2 rank masks + the shape ints) to a ``.npz``.

   Shares :py:func:`~t3toolbox.backend.common.save_core_families`: family 0 is the supercores, family 1
   is the (numpy, host) rank masks, family 2 is the static ``shape`` as a 1-element int array.
   :py:func:`ut3_load` regroups them. ``np.savez`` stores the boolean masks with their dtype, so
   ``ut3_load`` recovers numpy bool masks.